Académique Documents
Professionnel Documents
Culture Documents
FOREST SERVICE
RESEARCH PAPER
FPL-87
JUNE 1968
U. S. DEPARTMENT OF AGRICULTURE FOREST SERVICE FOREST PRODUCTS LABORATORY MADISON, WISCONSIN
1
FOREST PRODUCTS LABORATORY
FOREST SERVICE
INTRODUCTION
The solutions of many plane stress problems are literally impossible when attempted by applying
the differential equations of the theory of elasticity. For the solution of such problems, stress analysts
have sought other methods. One of these methods has been termed the finite element technique, and
appears to have merit in yielding approximate solutions to such problems.
The basic concept of the method consists of replacing the solid elastic body to be analyzed by a
network of finite elements. It is believed that as the size of the finite element approaches the differ-
ential element stage, the results yielded by the method would compare favorably to those obtained
from a rigorous mathematical analysis. By keeping the element finite in size, the network model
would no longer yield equivalent results but should represent a close approximation.
The finite element technique of plane stress analysis has been presented in different papers by
2
Hrennikoff (2), McCormick (4), Turner (6), and Melosh (5) to name a few. The technique has gained
considerable recognition with application to problems associated with the aircraft industry. In all of
these papers, however, the technique has been applied to problems associated with isotropic materials.
In general, the technique can be divided into two separate subcategories: (1) The framework method
and (2) the stiffness element method. The two methods differ principally in the composition of the
finite element. It is the purpose of this paper to examine each of these methods closely and determine
their applicability in handling problems in orthotropic plane stress.
Framework Method
The framework method is appropriately named since the method consists of replacing the solid
elastic body to be analyzed by a mathematical model of an imaginary framework. The framework is
composed of a series of pin-connected bars arranged in a definite pattern compatible with the of
problem solved, such as the cantilever beam of figure 1.
The most common pattern used in the framework method, and the one used by McCormick, is a
rectangular element diagonals connecting the corners, as shown in figure 2.
The areas of the bars A in the network system are determined by the requirement of equal deforma-
i
bility of the solid elastic body and the framework model under a given stress situation. Computations
for isotropic materials have been carried out by Hrennikoff and McCormick. In the process of calcula-
tion, however, it was discovered that a restriction must be imposed such that
FPL 87 2
where and are the two Poisson's ratios associated with the x-y plane and where , , and
yx xy x y
e are the strains associated with the x, y, and x-y directions, respectively. This might imply an even
xy
greater restriction of the framework method when applied to orthotropic materials. Computations
will now be carried out similar to those performed for isotropic material by Hrennikoff and McCormick,
to determine the restriction in orthotropic plane stress.
Mathematical analysis.--Consider a solid orthotropic elastic element (wood) of thickness t, and the
corresponding framework model subjected to a situation of pure shear stress:
where:
(2)
(3)
Also:
(4)
equating(3) and(4):
(5)
(6)
(7)
FPL 87 4
Consider now the stress situation existing such a s to create a case of strain in the x-direction
only, i.e.:
(8)
(9)
(10)
where
A pictorial representation of this stress situation would look like:
(11)
(12)
Therefore:
(13)
(14)
(15)
FPL 87 6
Therefore:
(16)
since
(17)
(18)
The same expression for S in (18) could be obtained using a stress situation such that
d
= 0. The point is, however, that it is also the function of the diagonal members to provide
stiffness for the Poissons effect when the model i s subjected to extensional strain only. The stiffness
(S ) a s given by (18) and (7) must be equal. Hence, equating, we have:
d
which in general i s not true for orthotropic materials. In view of this basic ingenerality, therefore,
it. is concluded that the framework method is not applicable to general orthotropic plane stress prob-
lems, but should give good results for the special case of orthotropy a s defined by equation (19).
The stiffness element method differs from that of the framework method in that the elements in the
network system are solid or plate elements, and further the elastic properties of the element should
duplicate the material it replaces. The elements still remain connected to each other only at the
corners or nodes, This perhaps can be visualized more easily if reference is made to a typical ortho-
tropic beam problem shown in figure 3 which will also be the subject of the following discussion on
orthotropic plane stress. From figure 3 it is easy to visualize that as the element size decreases, or
the number of elements increases, the behavior of the beam model will tend to approach the true
behavior of the orthotropic beam.
The stress situation in the beam model will be determined by the manner in which the forces are
propagated from element node to element node. For any one element the forces are related to the
displacement of the element nodes, since deformability is the physical feature determining the manner
of stress propagation through the stressed medium, Each node of the network may have external
forces applied in the x and y directions (coordinate system as shown in fig. 4). The deformation of
a single element, of thickness t, such as shown in figure 4 is defined by the eight possible nodal
displacements.
The relationships between the forces and displacements of a single element can be conveniently
handled in matrix form as:
(20)
SPECIFIC OBJECTIVES
The remaining objectives of this paper, therefore, are to: A. Develop the stiffness [K] matrix for a
single orthotropic element. B. Determine a technique in which the component stiffness matrices can
be conveniently handled without overflowing available computer capacity, C. Check these results
by investigating simple problems in which the stress distributions are well established. D. Develop
87 8
a stiffness [ K] matrix which might be used in orthotropic plate analysis. And finally, E. Check the
results of the bending stiffness matrix by analyzing a hypothetical orthotropic plate and comparing
the results with a rigorous mathematical analysis.
Consider for analysis the orthotropic plate element as shown in figure 4 with the forces and dis-
placements shown in their positive directions, The stresses and strains for such an element are related
by:
(21)
The coefficients in any one column of the matrix represent physically the forces which must be
applied at the nodes in order to give a displacement of unity for the particular column chosen while
the remaining displacements remain zero. It is this part of the derivation which determines the number
of strain expressions or alternatively the number of applied stress states which must be used to
achieve this. The number is always twice the number of nodes minus three. Hence, for a rectangular
element, five states are required, which will be:
For ease in derivation, the of superposition will be utilized so that each case may be handled
separately and the results added to determine the combined effect.
Consider first:
From equations (21), the stresses for an orthotropic element a r e found to be:
It is assumed, however, that these stresses can be replaced by the following equivalent forces
applied at the nodal points (refer to fig. 4 for sign convention):
Substituting, therefore:
Also:
or
Consider next:
FPL 87 10
where:
Then:
Also:
11
In equivalent force s y s t e m
and also:
Similarly:
and
Since
therefore
In considering the equivalent force system for this case, care must be taken to insure equilibrium.
The typical force relationship becomes:
FPL 87 12
Therefore:
or
where:
Then:
13
Since
therefore
The equivalent force system can be written:
where
Then:
By superimposing the relationships obtained between forces and displacements for the five stress
states, the final equations can be written in matrix form as:
(22)
where:
FPL 87 14
or simply:
Therefore:
15
where
With the establishment of the component stiffness matrix it is now desirable to formulate a composite
stiffness matrix which would relate, for example, all the nodal forces and deformations occurring in
our beam model in figure 3. For many problems, however, it is not necessary to model the whole
structure where in most places the stress distribution is well established, but only adjacent to loca-
tions where irregularities occur which might effect the normal stress distribution, For purposes of
verification and discussion, therefore, let us assume it is desirable to model a section of the beam
adjacent to the where symmetry might be observed, and use can be made of the boundary condition
that horizontal displacement at the must be zero. The beam model of such a section is shown in
figure 6.
FPL 87 16
The size of a network (9 x 9) of elements was limited to the available computer capacity. The
elements are numbered vertically and the arrows indicate the direction of positive forces and dis-
placements,
The composite stiffness matrix, then, will relate all the forces to displacements occurring in the
modeled section. Such a matrix, however, would represent the solution of 200 simultaneous equations,
which might prove too cumbersome for many small digital computers. It is necessary, therefore, to
develop a technique by which the matrix size might be kept small.
It is first necessary to build a typical stiffness matrix for the column of elements being modeled.
Since the model is assumed to be homogeneous, all such stiffness matrices are identical, the only
change being the nodal values. The stiffness matrix, for example, for the first column of elements can
be obtained by combining the elemental component matrices one through nine (see fig. 6). This results
in a matrix of size 40 x 40, relating the first 40 forces and displacements. The second matrix is
identical to the first with the exception it relates the forces and displacements 21 through 60. The
composite matrix for the first two columns of elements can he obtained by combining the individual
matrices, realizing that forces and displacement 21 through 40 are common to both columns of ele-
ments, resulting in a matrix of size 60 by 60. This may be represented in the form of a partitioned
matrix as:
(23)
(24)
17
To make use of (24), it is first convenient to interchange rows and columns of equations (23)
so to read:
(25)
or
(26)
where:
FPL 87 18
or
(27)
The forces and displacements 1 through 20 and 41 through 60 can now be written:
(28)
or finally:
(29)
19
This process of eliminating the common nodal values can be repeated until the desired network size
is achieved. For the particular size network chosen here, the final matrix is modified so that the
final equations become
(30)
With the establishment of the A modified matrix for the 9 by 9 network, it is now possible to check
the models ability to duplicate the original beam by subjecting it to stress distributions which are
well established and calculate whether the matrix analysis yields equivalent results. For purposes of
comparison, therefore, the following three loading conditions will be applied to the beam model in
figures 3 and 6; (1) a force F acting alone, such that a tensile stress of = 9,000 pounds per square
inch is induced, (2) the forces P acting alone such that a maximum bending stress of = 10,000 pounds
per square inch is induced in the extreme fibers of the beam model, and (3) a combination (1) and (2).
The given stress loading conditions will be applied as a statically equivalent force system derived
from states of loading designated as (a) and (d) in the section Develop K Matrix for Single Element.
These forces will be applied to the extreme left section of the beam model and then the stress situation
calculated at the centroid of each element in the ninth column. The stress distributions will then be
compared with the known distribution.
This operation can perhaps be seen more clearly by beginning with the equations:
(30)
FPL 87 20
yielding
(31)
The { f } matrix now represents the external force matrix which, for the three loading conditions,
i
is presented on page 22, where (dt) represents the cross-sectional area of the beam.
The displacements in equation (31) can now be determined by carrying out the matrix operation
defined, for each of the three cases. The displacement 6 through 6 can be gotten from node
161 180
elimination equation (27), which for addition of the ninth column would appear as
(where the D and C matrices are determined when the ninth column was added.)
21
FPL 87 22
With the solution of equations (31), all the nodal displacements for column 9 are known. It is now
a relatively easy operation to relate these displacements to corresponding stresses at the centroids
of each element, by making use of the equations already established in the section Develop K Matrix
for Single Element.
It is found that the stress matrix for any one element of such a 9 element column can be written:
where:
where the matrix represents the eight nodal displacements about the element maintaining coordinate
values similar to those in figure 4.
The resulting computed stress distributions for column 9 are presented in figure 7. On the basis of
this figure, therefore, it is concluded that the finite element technique was successful in modeling
these three simple cases of orthotropic plane stress.
23
Finite Element Technique in Orthotropic
Plate Analysis
Another area in which the finite element technique has gained popularity is in bending problems of
thin plates. Most of these problems again have been associated with the aircraft industry, where it
seems most of this work has been pioneered. Melosh (5), derives an elemental stiffness matrix
utilizing the bending strain energy expression for a uniform flexurally rigid isotropic plate. Most of
the derivations performed by Melosh were based on purely geometrical considerations, so that a
transformation to derive a plate stiffness matrix for orthotropic thin plates can be accomplished
rather easily. For this section, therefore, an elemental stiffness matrix will be derived for orthotropic
thin plates, for which the procedure will parallel that of Melosh with the exception of maintaining
orthotropic behavior,
Mathematical analysis.--It will be assumed that a given rectangular orthotropic (wood or plywood)
plate, as shown in figure 8a, can be modeled by an assemblage of rectangular elements connected at
their nodal points, as in figure 8b.
Each element in figure 8 b is assigned a stiffness matrix relating the forces and displacements at
its nodal points. Each node will have three degrees of freedom, the angles of rotation and about the
x and y axis, respectively, and the lateral displacement of each node, w . Associated with these dis-
i
placements are the forces at each node, respectively. The stiffness matrix for each
element will be of size 12 by 12. The stiffness matrix of the complete structure can then be determined
by addition of the individual component matrices.
The bending strain energy expression for a plate in which the axes of orthotropy
coincide with the plate boundaries is given by (1):
(32)
where
FPL 87 24
The stiffness matrix will be developed by adding matrices reflecting the stiffness of each term in
the bending energy expression, since each of these terms can be treated separately.
In deriving the plate stiffness matrix, it will be assumed that the bending curvature along the edges
of the plate can be expressed in terms of a third-order polynominal, or
(33)
where w now defines the displacements; for example, along edge 1-3 in Figure 9.
The four constants in expression (33) can be determined by applying the boundary conditions which
must exist along that edge; for example, along edge 1-3, the conditions are:
25
From these expressions, a relationship for along 1-3 can be written in matrix notation as:
(34)
The total strain energy resulting from can be found by squaring expression (34) and
integrating the result over half the plate area. (It is also assumed that expression (33) varies uniformly
with y.) The force-displacement relationships can then be found for this portion of the plate by utilizing
Castiglianos complementary relation:
(35)
3 The s u b s c r i p t o n t h e d i f f e r e n t i a l t e r m s d e n o t e s t h e edge a l o n g w h i c h t h e r e l a t i o n s h i p
applies.
FPL 87 26
A similar set of expressions can be found by treating edge 2-4 and using the remaining half of the
plate. By adding the two results, the final bending stiffness matrix due to the
equation (32) can be written:
(36)
27
If it is assumed that displacements along the edge in the Y-direction are of the same general form
as given by expression (33), then the bending stiffness matrix for the term can be written
by observing symmetry as:
(37)
To obtain the bending strain energy due to the bending coupling effect, it is necessary to perform
the integration:
over the area of the plate. This can be accomplished, as outlined by Melosh, by taking the product,
i.e., at each node and integrating over only a quadrant of the plate at a time. The sum of the
integral of the four products then represents the coupling energy. For example, to obtain the energy
for the first quadrant, it is necessary to perform the integration:
FPL 87 28
29
Treating the remaining three quadrants in a similar manner and superimposing the results lead to
the following stiffness matrix for the coupling energy t e r m of equation (32):
In deriving the stiffness matrix for the torsion t e r m of equation (32), it is necessary to determine
its corresponding energy or:
(39)
An expression for can be found by considering the simple torsion of the finite element of
figure 10, where:
(40)
FPL 87 30
By squaring equation (40), performing the integration of (39), and differentiation as done previously,
it can be found that the stiffness matrix for the torsion term can be written:
(41)
The final stiffness matrix for orthotropic thin plates can now be determined by superposition of
expressions(36), (37), (38), and (41), and found to be:
(42)
31
where:
Where for plywood plates, these elastic properties represent their effective values in the direction
denoted by the subscripts. These can be determined experimentally or by knowing the properties of
the individual plies, they can be calculated by existing formula (7).
With the establishment of the stiffness matrix for orthotropic thin plates, it is now desirable to
compare the finite element models ability to duplicate the behavior of an orthotropic plate as described
by a rigorous mathematical solution such as that derived by March (3).
For purposes of comparison, therefore, it will be assumed it is desirable to determine the
maximum deflection of the simply-supported five-ply plywood plate of Figure 10 under (1) a concen-
trated center load and (2) a uniformly distributed load
FPL 87 32
From March it can be found after considerable calculation that the maximum center deflection W is
0
given by:
for the case of the uniformly distributed load, where D is as previously defined.
To obtain comparative values by the finite element technique, the plate of Figure 10 will be modeled
by 36 finite elements as shown in Figure
33
By adding the individual elemental stiffness matrices it can be seen that the final composite matrix
will be of size 147 by This can be reduced, however, to a 48 by 48 matrix by observing symmetry
and analyzing only one quadrant for the particular cases chosen. By applying the appropriate boundary
conditions and inverting the resulting matrix, the maximumdeflections by the finite element model was
found to be:
and
FPL 87 34
CONCLUSIONS
On the basis of the limited examples examined in this report, it appears the finite element technique
has merit regarding problems in orthotropic plane stress. It is realized that the networks used in this
report were extremely coarse, but this should not distract the fact of the models ability to duplicate
orthotropic behavior. It is felt that the techniques which need to be developed in the future lie in the
effective handling of the large systems of equations that result a s finer networks a r e desired. The
primary obstacle possibly is the round-off e r r o r occurring in the many digital computations.
35
LITERATURE CITED
1. Hearmon, R.F.S.
1961. Applied anistropic elasticity. Oxford University Press.
2. Hrennikoff, A,
1941. Solutions of problems of elasticity by the framework method. J. Appl. Mech. 8(4):A169-
A175 (Dec.)
3. March, H.W.
1942. Flat plates ofplywoodunder uniformor concentrated loads. Forest Products Lab. Rep. 1312.
4. McCormick, C.
1963. Plane stress analysis. J. Struc. Div., Proc. Amer. Soc. Civil Eng. (Aug.)
5. Melosh, R. J.
1961. A stiffness matrix for the analysis ofthinplates in bending. J. Aero. Sci. 28(1):34-42 (Jan.)
FPL 87 36
APPENDIX 1
S t r a i n i n t h e x- D i r e c t i o n o n l y .
37
S t r a i n i n t h e y- D i r e c t i o n o n l y .
Shear S t r a i n i n x - y Plane.
FPL 87 38
NOTATION
Strain (shear); the change in angle between lines originally drawn in the x and y directions.
Modulus of rigidity associated with shear deformation in the x-y plane resulting from shear
stresses in the xz and yz planes.
Poisson's ratio of contraction along the direction y to extension along the direction x due to
a normal tensile stress in the direction x; similarly,
The stiffness values (E A ) of the diagonal and horizontal member in the framework model.
i i
39
FPL 87 40 1.5-40